ai-fine-tuning
Pass
Audited by Gen Agent Trust Hub on May 1, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows security best practices and provides legitimate educational content for AI development.- [EXTERNAL_DOWNLOADS]: Provides commands to install additional utility skills from the same author using 'npx skills add'. These references are part of the intended developer workflow for the DSPy ecosystem.- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates how to ingest and process local JSON datasets (e.g., 'labeled_data.json', 'tickets.json') for fine-tuning. While this is an ingestion surface for potentially untrusted data, it is a necessary part of the fine-tuning process.
- Ingestion points: SKILL.md (line 92), examples.md (lines 28, 102, 175, 239).
- Boundary markers: Absent; the skill assumes user-provided labeled datasets are trusted.
- Capability inventory: Employs dspy optimizers (BootstrapFinetune, MIPROv2) to generate model traces and execute fine-tuning tasks.
- Sanitization: Uses standard JSON loading without explicit filtering of data content.
Audit Metadata